langchain-ocr-lib 0.2.0__tar.gz → 0.3.1__tar.gz

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/PKG-INFO

@@ -1,7 +1,7 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: langchain-ocr-lib
-Version: 0.2.0
-Summary: 
+Version: 0.3.1
+Summary: Modular, vision-LLM-powered chain to convert image and PDF documents into clean Markdown.
 License: MIT
 Author: Andreas Klos
 Author-email: aklos@outlook.de
@@ -54,7 +54,7 @@ This package offers the core functionality to extract text from documents using
 
 ## 2. Features
 
-- **Vision-Language OCR:** Supports Ollama. Other LLM providers will be added soon.
+- **Vision-Language OCR:** Supports Ollama, vLLM and OpenAI (and other OpenAI conform providers). Other LLM providers can be easily integrated.
 - **CLI Interface:** Simple local execution via command line or container
 - **Highly Configurable:** Use environment variables to configure the OCR
 - **Dependency Injection:** Easily swap out components for custom implementations
@@ -70,7 +70,7 @@ This package offers the core functionality to extract text from documents using
 - **Python:** 3.11+
 - **Poetry:** [Install Poetry](https://python-poetry.org/docs/)
 - **Docker:** For containerized CLI usage (optional)
-- **Ollama:** Follow instructions [here](https://ollama.com)
+- **Ollama:** Follow instructions [here](https://ollama.com) (other LLM providers can be used as well, see [here](#2-features))
 - **Langfuse:** Different options for self hosting, see [here](https://langfuse.com/self-hosting) (optional, for observability)
 
 ### 3.2 Environment Setup

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/README.md

@@ -27,7 +27,7 @@ This package offers the core functionality to extract text from documents using
 
 ## 2. Features
 
-- **Vision-Language OCR:** Supports Ollama. Other LLM providers will be added soon.
+- **Vision-Language OCR:** Supports Ollama, vLLM and OpenAI (and other OpenAI conform providers). Other LLM providers can be easily integrated.
 - **CLI Interface:** Simple local execution via command line or container
 - **Highly Configurable:** Use environment variables to configure the OCR
 - **Dependency Injection:** Easily swap out components for custom implementations
@@ -43,7 +43,7 @@ This package offers the core functionality to extract text from documents using
 - **Python:** 3.11+
 - **Poetry:** [Install Poetry](https://python-poetry.org/docs/)
 - **Docker:** For containerized CLI usage (optional)
-- **Ollama:** Follow instructions [here](https://ollama.com)
+- **Ollama:** Follow instructions [here](https://ollama.com) (other LLM providers can be used as well, see [here](#2-features))
 - **Langfuse:** Different options for self hosting, see [here](https://langfuse.com/self-hosting) (optional, for observability)
 
 ### 3.2 Environment Setup

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/pyproject.toml

@@ -1,5 +1,5 @@
 [build-system]
-requires = ["poetry-core==1.8.5"]
+requires = ["poetry-core==2.1.2"]
 build-backend = "poetry.core.masonry.api"
 
 [tool.poetry.scripts]
@@ -7,8 +7,8 @@ langchain-ocr = "langchain_ocr_lib.main:main"
 
 [tool.poetry]
 name = "langchain-ocr-lib"
-version = "0.2.0"
-description = ""
+version = "0.3.1"
+description = "Modular, vision-LLM-powered chain to convert image and PDF documents into clean Markdown."
 authors = ["Andreas Klos <aklos@outlook.de>"]
 readme = "README.md"
 packages = [{ include = "langchain_ocr_lib", from = "src" }]
@@ -59,7 +59,7 @@ flake8-docstrings = "^1.7.0"
 
 
 [tool.flake8]
-exclude= [".eggs", ".git", ".hg", ".mypy_cache", ".tox", ".venv", ".devcontainer", "venv", "_build", "buck-out", "build", "dist", "**/__init__.py", "src/langchain_ocr_lib/prompt_templates/ocr_prompt.py"]
+exclude= [".eggs", ".git", ".hg", ".mypy_cache", ".tox", ".venv", ".devcontainer", "venv", "_build", "buck-out", "build", "dist", "**/__init__.py", "src/langchain_ocr_lib/prompt_templates/ocr_prompt.py", "tests"]
 statistics = true
 show-source = false
 max-complexity = 8

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/src/langchain_ocr_lib/di_config.py

@@ -14,9 +14,11 @@ from langchain_ocr_lib.di_binding_keys.binding_keys import (
 from langchain_ollama import ChatOllama
 from langchain_openai import ChatOpenAI
 from langfuse import Langfuse
+from functools import partial
 
 from langchain_ocr_lib.impl.chains.ocr_chain import OcrChain
 from langchain_ocr_lib.impl.settings.ollama_chat_settings import OllamaSettings
+from langchain_ocr_lib.impl.settings.vllm_chat_settings import VllmSettings
 from langchain_ocr_lib.impl.settings.openai_chat_settings import OpenAISettings
 from langchain_ocr_lib.impl.settings.llm_class_type_settings import LlmClassTypeSettings
 from langchain_ocr_lib.impl.settings.langfuse_settings import LangfuseSettings
@@ -49,13 +51,17 @@ def lib_di_config(binder: Binder):
 
     if llm_class_type_settings.llm_type == "ollama":
         settings = OllamaSettings()
-        llm_instance = llm_provider(settings, ChatOllama)
+        partial_llm_provider = partial(llm_provider,settings, ChatOllama)
     elif llm_class_type_settings.llm_type == "openai":
         settings = OpenAISettings()
-        llm_instance = llm_provider(settings, ChatOpenAI)
+        partial_llm_provider = partial(llm_provider,settings, ChatOpenAI)
+    elif llm_class_type_settings.llm_type == "vllm":
+        settings = VllmSettings()
+        partial_llm_provider = partial(llm_provider,settings, ChatOpenAI)
     else:
         raise NotImplementedError("Configured LLM is not implemented")
-    binder.bind(LargeLanguageModelKey, llm_instance)
+    
+    binder.bind_to_provider(LargeLanguageModelKey, partial_llm_provider)
 
     prompt = ocr_prompt_template_builder(language=language_settings.language, model_name=settings.model)
 

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/src/langchain_ocr_lib/impl/converter/pdf_converter.py

@@ -48,9 +48,11 @@ class Pdf2MarkdownConverter(File2MarkdownConverter):
                 with open(filename, "rb") as f:
                     file = f.read()
             except Exception as e:
-                raise ValueError("PDF corrupted or unsupported file type, %s" % e)
-
-        images = convert_from_bytes(file)
+                raise ValueError("PDF corrupted or unsupported file type") from e
+        try:
+            images = convert_from_bytes(file)
+        except Exception as e:
+            raise ValueError("PDF corrupted or unsupported file type") from e
 
         markdown = ""
         for image in images:
@@ -93,7 +95,10 @@ class Pdf2MarkdownConverter(File2MarkdownConverter):
             except Exception as e:
                 raise ValueError("PDF corrupted or unsupported file type") from e
 
-        images = convert_from_bytes(file)
+        try:
+            images = convert_from_bytes(file)
+        except Exception as e:
+            raise ValueError("PDF corrupted or unsupported file type") from e
 
         markdown = ""
         for image in images:

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/src/langchain_ocr_lib/impl/langfuse_manager/langfuse_manager.py

@@ -35,20 +35,6 @@ class LangfuseManager:
     ):
         self._managed_prompts = managed_prompts
 
-    def init_prompts(self) -> None:
-        """
-        Initialize the prompts managed by the LangfuseManager.
-
-        This method iterates over the keys of the managed prompts and retrieves
-        each prompt using the `get_langfuse_prompt` method.
-
-        Returns
-        -------
-        None
-        """
-        for key in list(self._managed_prompts.keys()):
-            self.get_langfuse_prompt(key)
-
     def get_langfuse_prompt(self, base_prompt_name: str) -> Optional[ChatPromptClient]:
         """
         Retrieve the prompt from Langfuse Prompt Management.

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/src/langchain_ocr_lib/impl/llms/llm_type.py

@@ -9,3 +9,4 @@ class LLMType(StrEnum):
 
     OLLAMA = "ollama"
     OPENAI = "openai"
+    VLLM = "vllm"

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/src/langchain_ocr_lib/impl/settings/ollama_chat_settings.py

@@ -35,7 +35,7 @@ class OllamaSettings(BaseSettings):
         env_prefix = "OLLAMA_"
         case_sensitive = False
 
-    model: str = Field(default="gemma3:4b-it-q4_K_M")
+    model: str = Field(default="gemma3:4b-it-q4_K_M", title="LLM Model")
     base_url: str = Field(default="http://localhost:11434")
     top_k: int = Field(default=0, title="LLM Top K")
     top_p: float = Field(default=0, title="LLM Top P")

{langchain_ocr_lib-0.2.0 → langchain_ocr_lib-0.3.1}/src/langchain_ocr_lib/impl/settings/openai_chat_settings.py

@@ -18,8 +18,8 @@ class OpenAISettings(BaseSettings):
         Total probability mass of tokens to consider at each step.
     temperature : float
         What sampling temperature to use.
-    vision_capable : bool
-        Flag to enable a vision capable model.
+    base_url : str
+        The base URL for the OpenAI API endpoint.
     """
 
     class Config:
@@ -28,8 +28,11 @@ class OpenAISettings(BaseSettings):
         env_prefix = "OPENAI_"
         case_sensitive = False
 
-    model: str = Field(default="gpt-4o-mini-search-preview-2025-03-11", description="The model identifier")
+    model: str = Field(default="gpt-4o-mini-search-preview-2025-03-11", description="The model identifier", title="LLM Model")
     api_key: str = Field(default="", description="The API key for authentication")
-    top_p: float = Field(default=1.0, description="Total probability mass of tokens to consider at each step")
-    temperature: float = Field(default=0.7, description="What sampling temperature to use")
-    vision_capable: bool = Field(default=False, description="Enable a vision capable model")
+    top_p: float = Field(default=1.0, description="Total probability mass of tokens to consider at each step", title="Top P")
+    temperature: float = Field(default=0, description="What sampling temperature to use", title="Temperature")
+    base_url: str = Field(
+        default="https://api.openai.com/v1",
+        description="The base URL for the OpenAI API endpoint",
+    )

langchain_ocr_lib-0.3.1/src/langchain_ocr_lib/impl/settings/vllm_chat_settings.py

@@ -0,0 +1,38 @@
+"""Module contains settings regarding the Vllm API."""
+
+from pydantic import Field
+from pydantic_settings import BaseSettings
+
+
+class VllmSettings(BaseSettings):
+    """
+    Contains settings regarding the Vllm API.
+
+    Attributes
+    ----------
+    model : str
+        The model identifier.
+    api_key : str
+        The API key for authentication.
+    top_p : float
+        Total probability mass of tokens to consider at each step.
+    temperature : float
+        What sampling temperature to use.
+    base_url : str
+        The base URL for the Vllm API endpoint.
+    """
+
+    class Config:
+        """Config class for reading fields from environment variables."""
+
+        env_prefix = "VLLM_"
+        case_sensitive = False
+
+    model: str = Field(default="", description="The model identifier", title="LLM Model")
+    api_key: str = Field(default="", description="The API key for authentication")
+    top_p: float = Field(default=1.0, description="Total probability mass of tokens to consider at each step", title="Top P")
+    temperature: float = Field(default=0, description="What sampling temperature to use", title="Temperature")
+    base_url: str = Field(
+        default="http://localhost:8000/v1",
+        description="The base URL for the Vllm API endpoint",
+    )